aiqbench analysis
This paper introduces a benchmark that tests whether LLM agents can build the model structure itself, not just fit parameters to a fixed template. Results are modest: nine models solve 21% to 60% of 80 simulated systems, and classical equation-discovery methods rarely succeed on most of them. For AI readers, the key finding is that the bottleneck is discovering the right structure, and agents tend to prefer revisions that lower training error over those that predict held-out behavior.
Unsatz: Evaluating Scientific Discovery Without a Given Ansatz
arXiv:2611.03713v1 Announce Type: cross Abstract: Before a scientist fits any parameter, they must decide what the model contains: which quantities exist, which are hidden, how they interact, and how they are measured. Physicists call such a trial form an ansatz. Benchmarks for automated equation discovery mostly supply it, by fixing the variables, a library of candidate terms, or a law to be modified. We introduce Unsatz, a benchmark in which an LLM agent must construct the ansatz itself. Each task is a simulated system, generated from a private seed as a program in a small modeling language: ordinary, delay, or partial differential equations, discrete maps, static and causal models, or stochastic processes, possibly with hidden states, distorted sensors, and uninformative measurement channels. The agent buys experiments within a budget, writes candidate programs in the same language, obtains their coefficients from the benchmark's fitter, and is scored only on how its final program predicts experiments it never saw: conditions outside the range it could explore, interventions, and longer runs. Across 80 systems in 20 domains, nine language models solve between 21% and 60%. On 44 of these systems, classical equation-discovery methods almost never succeed, although the same data usually suffice once the true equations are given with unknown coefficients; the difficulty lies in finding the structure, not in estimating it. Because mechanisms recur across systems, we also give agents the program they discovered for a related system; it helped only when the new system shared the mechanism and the first investigation had succeeded. Stronger agents make fewer unscientific errors; beyond these, agents find which variables a system has but not how they act, and prefer revisions that lower the training error to revisions that predict better.
Abstract:Before a scientist fits any parameter, they must decide what the model contains: which quantities exist, which are hidden, how they interact, and how they are measured. Physicists call such a trial form an ansatz. Benchmarks for automated equation discovery mostly supply it, by fixing the variables, a library of candidate terms, or a law to be modified. We introduce Unsatz, a benchmark in which an LLM agent must construct the ansatz itself. Each task is a simulated system, generated from a private seed as a program in a small modeling language: ordinary, delay, or partial differential equations, discrete maps, static and causal models, or stochastic processes, possibly with hidden states, distorted sensors, and uninformative measurement channels. The agent buys experiments within a budget, writes candidate programs in the same language, obtains their coefficients from the benchmark's fitter, and is scored only on how its final program predicts experiments it never saw: conditions outside the range it could explore, interventions, and longer runs. Across 80 systems in 20 domains, nine language models solve between 21% and 60%. On 44 of these systems, classical equation-discovery methods almost never succeed, although the same data usually suffice once the true equations are given with unknown coefficients; the difficulty lies in finding the structure, not in estimating it. Because mechanisms recur across systems, we also give agents the program they discovered for a related system; it helped only when the new system shared the mechanism and the first investigation had succeeded. Stronger agents make fewer unscientific errors; beyond these, agents find which variables a system has but not how they act, and prefer revisions that lower the training error to revisions that predict better.
Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2611.03713 [cs.AI]
(or arXiv:2611.03713v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2611.03713 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 377 words
Related coverage
Style Wins, Substance Loses: A Diagnosis of LLM-as-Judge in Idea Generation
arXiv · Aug 4, 2026
ResearchClawBench: A Benchmark for End-to-End Autonomous Scientific Research
arXiv · Jun 12, 2026
ResearchClawBench: A Benchmark for End-to-End Autonomous Scientific Research
arXiv · Jun 9, 2026
SciDiagBench: A Type-Specific Benchmark for Human-Aligned Evaluation of Scientific Diagrams
arXiv · Oct 21, 2026